Numerical calculation of the wind action on buildings using Eurocode 1 atmospheric boundary layer velocity profiles

Transcription

1 Numerical calculation of the wind action on buildings using Eurocode 1 atmospheric boundary layer velocity profiles M. F. P. Lopes IDMEC, Instituto Superior Técnico, Av. Rovisco Pais 149-1, Lisboa, Portugal mlopes@hidro1.ist.utl.pt Introduction In the design of the aerodynamic behaviour of a building or structure, there are several levels of insight that can be considered. Both the static forces and the dynamic effects caused by the wind action can be relevant to the design of the structure. For the calculation of these effects in specific cases where the information provided in Eurocode 1.4 [1] in not enough, the choice between wind tunnel testing and numerical calculations using Computational Fluid Dynamics (CFD) codes is still not straightforward []. The main difficulties in calculating the wind action on buildings using CFD are [3]: high Reynolds number, impinging at the front face, sharp edges of bluff bodies and effect of the wake in the outflow boundary. In the present case, where it is also aimed to analyse the effect of the incident velocity profile, the problem of the maintenance of the ABL velocity profile is also conditioning. To the calculation of the wind action on buildings it is first needed to define the characteristics of the mean velocity and turbulence profiles. This definition is frequently not precise due to the non uniform conditions of aerodynamic roughness, caused by other buildings and the topography characteristics. As a consequence, and to take into consideration the effect of the neighbouring structures and simplify the design process, it is considered that the variation of the mean velocity and turbulence is represented by profiles that are only dependent on the height and a roughness parameter. In the EC1[1], the mean velocity profiles are defined through a logarithmic formulation:,7 z z U ( z) = U,19 ln b, (1),5 z where U b is the design velocity, characteristic of each geographic area; z is the height above the ground and z is the aerodynamic roughness length. For a height lower than z min, the mean velocity is considered constant, with the value correspondent to z min. This aims to take into account the unknown variability of the wind velocity in the first meters above the ground. Equation 1 can be related with the classical definition of the logarithmic profile through a parameter d : 1

2 z u * d = U b.19 =, ().5 K where, u * is the friction velocity and K is the Von Karman constant (approximately.4) The EC1 defines five terrain categories, according to their aerodynamic roughness, corresponding to the parameters represented in Table 1:.7 Table 1 Terrain categories according to EC1 [1] Terrain Category z [m] z min [m] Sea or coastal area exposed to open sea.3 1 I Lakes or flat and horizontal area with negligible vegetation and without obstacles II Area with low vegetation such as grass and isolated obstacles (trees, buildings) with separations of at least obstacle heights III Area with regular cover of vegetation or buildings or with isolated obstacles with separations of maximum obstacle heights IV Area in which at least 15% of the surface is covered with buildings and their average height exceeds 15 m,1 1,5,3 5 1, 1 The turbulence intensity, defined as the ratio between the standard deviation and the mean wind velocity, is defined as: 1 I u ( z) =. (3) ln ( z z ) The values of k (turbulent kinetic energy) and ε (dissipation rate of the turbulent kinetic energy) are usually obtained through the following expressions: ( U ( z). I ( z) ) 3 k = u, (4) 3 d K ε =. (5) z As a consequence of their definition, the EC1 ABL profiles correspond to constant turbulent kinetic energy profiles, only dependent on the terrain roughness. In this work, two groups of cases were analyzed. In the first group, only the upstream part of the numerical domain was used, without any obstacles. This was done to evaluate how a ABL velocity and turbulence distributions can be effectively maintained from the inflow boundary to the building. In a second stage, the flow in the complete domain and with the building was calculated using as inflow conditions either a uniform flow or the EC1 velocity and turbulence profiles. Model definition a) Mesh and domain specification The application case under analysis is a 3 m side cubic building. The following simplifications were assumed to build the numerical model: (a) the building has sharp eaves; (b) the building s walls are flat and without roughness elements (windows, balconies); (c) the aerodynamic effects of the

3 surrounding buildings are completely represented by the shape of the ABL profiles; (d) the wind direction is normal to one of the building walls. The numerical domain was defined according to the following dimensions relative to the cube s side L: in the direction of the flow L, being 9 L upstream and 1 L downstream the building; the domain s height was defined as 8 L and the width 1 L. The criteria for the specification of these dimensions were: in the inflow, up and side boundaries of the domain, to avoid obstacle influenced static pressure distributions on these boundaries; in the outflow, avoid the limitation of the wake s length to influence the correct representation of the flow. The mesh was built with two mesh grid blocks, taking into account the type of flow. In the zone around the cube, in a volume with dimensions 3 L 3 L L. (length, width, height), a structured mesh was defined with growing elements starting from the cube s walls. The thickness of the first element from the walls (. cm) was defined in order to accomplish the condition of having the values of the non-dimensional parameter + y in the range 3-3 (see [4] and the results section). Outside this inner domain an unstructured mesh was built, being finer near the floor and at the center of the domain after the obstacle, to account for the larger velocity gradients in these zones. Each of the building s faces was discretized by 5 5 elements. The final mesh was made of elements. b) Boundary Conditions The mean velocity, k and ε profiles were defined in the inflow boundary according to Equations (1) to (5) and Table 1 through a user defined function (UDF). The lateral and top boundaries of the domain were defined as solid walls with zero shear stress. The obstacle s walls were defined with a no slip condition. The application of the boundary condition on the floor implies special care, which is related to the maintenance of a ABL profile along a numeric domain. By definition, a ABL is in equilibrium with a uniformly distributed aerodynamic roughness on the floor. According to Blocken et al.[5], 4 conditions are needed to accomplish the equilibrium of an ABL flow type through an empty domain in a commercial code: (a) a sufficiently fine mesh in the vertical direction near the floor of the computational domain; (b) a horizontal homogeneous ABL in the upstream and downstream region of the domain; (c) a distance from the centre point of the wall adjacent cell to the bottom wall larger than the physical roughness height; (d) the knowledge of the relation between the sand-grain type roughness and the correspondent roughness height. However, it is not possible in the standard commercial codes to fulfill all the conditions, because of the definition of the wall functions (that are usually not-changeable). Computing a code using the model of Richards and Hoxey [6] however, it is possible to achieve the stability of the ABL in a very long domain. In FLUENT it is feasible to change the value of the floor s roughness. In FLUENT, this is made using the Standard Wall Functions and changing the parameters Roughness height ε R and Roughness constant C s. The value of ε R is different from z (roughness lenght). These values are related by (see demonstration in [5]): E ε R = z (6) C s 3

4 where E is a empirical constant whose value is in FLUENT. Using the standard value for.5, the following is obtained: C s of ε R = z (7). c) Turbulence model The choice of the adequate turbulence model is essential to solve the air flow around a building effectively. In this work, the turbulence models k ε standard and k ε with MMK modification were used. The necessity of modifying the standard model arises from the excessive turbulent kinetic energy near the obstacle when this model is used in bluff bodies [3]. The modification known as MMK was first presented in [7]. This modification changes the calculation of the turbulent viscosity v t to reduce the production of turbulent kinetic energy around the body. Having S and Ω, respectively the strain and vorticity invariants: 1 du du i j S = +, (8) dx j dxi the following modification is introduced: Ω 1 du = dx i j du dx i j, (9) * k * Ω Ω * Ω vt = Ct, C t = C µ (se <1) ou C t = Cµ (se 1). (1) ε S S S This means, in practice, the reduction of the turbulence viscosity in the zones where S > Ω. In this work, this was applied by choosing the standard k ε turbulence model and introducing the UDF proposed in [3] to change the calculation of the turbulent viscosity. Results a) Verification of the ABL stability along an empty domain In order to guarantee that the ABL characteristic mean velocity and turbulent kinetic energy profiles are not significantly changed when passing an empty numerical domain, this problem was first addressed during the study. Figure 1 presents the results of the verification of the mean velocity profile maintenance along the domain that would later be upstream the building. The results were obtained using a velocity profile correspondent to the terrain type III of EC1 and a base velocity of 1 m/s, varying the floor boundary condition. The roughness was set to 6 m using Equation 7 and Table 1. The results show that the velocity variation relatively to the inflow velocity profile can be significant for the zones near the floor in this commercial code. It is also verified that adding a roughness condition in the floor boundary has a significant effect in helping the maintenance of the profile. In the outer region of the ABL, there is a slight increase of velocity, which is a consequence of momentum equilibrium. As verified from the 4

5 numerical data, this corresponds to an ascendant velocity gradient, specially significant in the beginning of the domain. 7 Height [m] , 3,4 Case 1 - Zero shear stress Case-Noslipw/smoothwall Case 3 - Roughness height 6 m Case 4 - Case 3 + MMK Original ABL profile Velocity variation (% of the original velocity) Figure 1 Verification of the outflow velocity after a 7 m numerical domain, relatively to the inflow velocity profile type III of EC1 (only represented from z min = 5 m) with several floor boundary conditions. U b = 1 m/s. Case 1 Zero shear stress; Case No slip with no roughness; Case 3 6 m Roughness Height; Case 4 6 m Roughness Height + MMK modification to the turbulence model. Figure presents the maintenance of the velocity profiles for the same numerical domain, varying the inflow velocity profile. From the obtained data it can be seen that the difficulty of maintaining of the inflow profile is more significant for the profiles correspondent to more rough terrains. This can be due to the inability of the sand-grain rough wall conditions in FLUENT to represent large roughness conditions, as discussed extensively in [8] and [5]. 7 Height [m] Terrain type zero Terrain type III Terrain type IV Original ABL profile Velocity variation (% of the original velocity) Figure Velocity variation at the end of the numerical domain, for Case 4 of Figure 1, changing the inflow velocity profile profiles for terrain types zero, III and IV are represented. The maintenance of the inflow kinetic energy profile was also verified and is represented in Figure 3. Due to the rough wall definition in FLUENT, that implies a peak in the turbulent kinetic energy dissipation in the cell adjacent to the walls, the maintenance of the turbulent kinetic is not well 5

6 accomplished. The case (a) in Figure 3, where the standard k - ε is used, was the one where best results were accomplished. In (b), where the MMK was added to the turbulence model, there is a higher variation of k, which may be related to the fact that the turbulent kinetic energy is limited when adding this alteration. (a) (b) Figure 3 Variation of the turbulent kinetic energy k along the numerical domain. Inflow constant distribution correspondent to terrain type III of EC1. Flow is in the positive yy axis direction. The inflow and outflow planes are represented, as well as the longitudinal mean plane. (a) Case 3; (b) Case 4. b) Verification of y + values The values of y + were computed during the mesh construction, in order to allow the usage of the standard wall functions. The non-dimensional parameter y + is used to check the correct representation of the object s boundary layer by the first layer of volumes next to it, and is defined as: y + * ρu y = µ P, (11) Here, ρ and µ are respectively the density and kinematic viscosity of the fluid in the first volume next to the wall and y P is the distance of the center of that element to the wall. The value of the thickness of the first element was changed in order to obtain the condition 3 < y + < 3, and a value of cm was obtained for the first element thickness. Figure 4 shows the values of + y on the center of the first layer of volumes that surround the cube. It can be observed that, with the exception of a small area in the front face, the indicative range is accomplished. 6

7 (a) (b) Figure 4 Values of y + on the center of the first layer of volumes that surround the cube. The flow has the direction of positive y. (a) windward, roof and side faces (b) leeward face. c) Pressure coefficients for uniform and ABL flows The pressure coefficients, i.e. the values of the dynamic pressure non-dimensionalized by the flow velocity in an undisturbed point at the obstacle height, were determined for uniform and ABL flows. From the analysis of Figure 5 and Figure 6, where the incident flow is uniform, it is noticeable that in the windward face the results represent well the experimental results and the difference between the two turbulence models is small. On the roof and the side faces, a significant deviation is verified between the experimental and numerical results near the separation eaves. The pressure coefficients obtained for an incident profile correspondent to the EC1 ABL velocity profiles are shown in Figure 7 e Figure 8. It can be noticed that the MMK modification is essential to solve the cases where the turbulence is high, as it is the case of the ABL-type flows. This becomes evident in the windward face as the values of C p are higher than 1 in this face without the MMK modification. Although the distribution in the central alignment of the front faces is satisfactory, the values near the side eaves and near the floor are slightly higher than expected. This may be due to the excessive velocity near the floor reported in Figure. The pressure distribution in the suction faces, near the separation eaves, as in the uniform case flow does not represent with precision the target values. This fact is also reported in [11] and [1] and is justified by the difficulty in reproducing the inversion of the flow direction inside the separation bubble near the suction faces. 7

8 Figure 5 Pressure Coefficients on the central alignments of the faces. Comparison between the models with and without the MMK modification and experimental results by Castro and Robbins [9] -.3 cp Figure 6 - C p values in the faces of the cube for uniform incident flow with the MMK modification of the turbulence model. 8

9 Figure 7 - Pressure Coefficients on the central alignment of the faces. Comparison between the models with and without the MMK modification and experimental results by Stathopoulos and Dumitrescu-Brulotte [1]. The distribution of pressure coefficients obtained for the ABL profiles adopted by the EC1 is represented in Figure 8 (a) to (d). The biggest differences between the results occur when passing from a uniform flow to the terrain type zero profile. With a uniform incident flow the stagnation point is located near the ground, whereas in the terrain type zero profile it is located at / 3 of the windward face height. The stagnation point is located increasingly higher for the profiles correspondent to more rough terrains, being located at about 3 / 4 for the terrain type III profile. It was not possible to obtain a realist pressure distribution correspondent the type IV terrain, due to the significant change of this profile between the inlet and the region near the obstacle, as seen in Figure. Conclusions In this paper, the numerical calculation of the wind action on buildings is addressed, focusing on the influence of the use of the mean velocity and turbulence profiles in the calculation. In the first set of results obtained in this work, where only the part of the numerical model upstream the obstacle was used, it was verified that the usage of the correct boundary condition on the floor is crucial. However, using the option of numerical roughness available in FLUENT has revealed to be insufficient to maintain the velocity profile, as the code is not prepared for large-scale aerodynamic roughness. To the calculation of the pressure coefficients in the cubic building s walls, it was necessary to change the definitions of the standard k -ε model applying the MMK alteration. It was verified that this was needed to prevent the high turbulence in the impinging region that caused coefficients higher than 1 in the windward face. However, it was noticed in the runs with a empty domain that this modification increased the dissipation of the ABL-characteristic k along the domain. With this modification, values closer to the target ones were obtained, although the flow inside the suction bubble near the sharp eaves was still not accurately represented. 9

10 cp (a) (b) cp (c) (d) Figure 8 - C p values in the faces of the cube for incident ABL profiles flow with the MMK modification of the turbulence model. (a) Terrain Type, (b) Terrain Type I (c) Terrain Type II; (d) Terrain Type III 1

11 References [1] Eurocode 1: Actions on structures General Actions Part 1-4: Wind Actions 4. [] T. Stathopoulos - The numerical wind tunnel for industrial aerodynamics: real or virtual for the new millennium? - Journal of Wind Engineering and Industrial Aerodynamics, 81, pp , [3] S. Huang, Q. S. Li, S. Xu, Numerical evaluation of wind effects on a tall steel building by CFD, Journal of Constructional Steel Research, 63, pp , 7. [4] FLUENT 6. User s Guide, USA, 5. [5] B. Blocken, T. Stathopoulos, J. Carmeliet - CFD simulation of the atmospheric boundary layer: wall function problems - Atmospheric Environment, 41, pp. 38-5, 7. [6] P.J Richards, R. Hoxey - Appropriate boundary conditions for computational wind engineering models using the κ-ε model - Journal of Wind Engineering and Industrial Aerodynamics, 46-47, pp , [7] M. Tsuchiya, S. Murakami, A. Mochida, K. Kondo, Y. Ishida; Development of new k-ε model for flow and pressure fields around bluff body; Journal of Wind Engineering and Industrial Aerodynamics, 67-68, pp , [8] D. M. Hargreaves, N.G.Wright - On the use of the κ-ε model in commercial CFD software to model the neutral atmospheric boundary layer - Journal of wind engineering and industrial aerodynamics, 95, pp , 7. [9] I. P. Castro, A.G. Robbins The Flow around a Surface-Mounted Cube in Uniform and Turbulent Streams Journal of Fluid Mechanics, 79, pp , [1] T. Stathopoulos, M. Dumitrescu-Brulotte Design Recommendations for Wind Loading on Structures of Intermediate Height Canadian Journal of Civil Engineering, 16, pp , [11] S. Murakami, A. Mochida, Y. Hayashi, S. Sakamoto - Numerical study on velocity-pressure field and wind forces for bluff bodies by k-ε, ASM and LES; Journal of Wind Engineering and Industrial Aerodynamics, 41-44, pp , 199. [1] M. Tsuchiya, S. Murakami, A. Mochida, K. Kondo, Y. Ishida; Development of new k-ε model for flow and pressure fields around bluff body; Journal of Wind Engineering and Industrial Aerodynamics, 67-68, pp ,